模分复用系统解复用技术研究

发布时间：2018-03-20 19:21

本文选题：模式分集复用　切入点：解复用　出处：《吉林大学》2016年博士论文　论文类型：学位论文

【摘要】：在全球数据流量爆发式增长的背景下,单模光纤通信系统容量已越来越接近其非线性香农极限。作为突破单模光纤容量极限的一种新的扩容技术,模式分集复用技术通过挖掘模式这一新的自由度可以成倍提升光纤通信系统容量。然而,在实际的基于少模光纤的模分复用系统中,模式耦合以及差分模式群时延等损伤会严重劣化模分复用系统的传输性能、限制模分复用系统的传输距离。在模式耦合和差分模式群时延的共同作用下,信号在模分复用系统中的传输过程要比单模光纤系统更为复杂:系统每路输出信号都是多路输入信号与信道冲激响应线性卷积的叠加。因此,如何有效补偿模式耦合和差分模式群时延等损伤,解决多信道卷积性混合问题,实现输出信号解复用是模分复用系统的关键问题。本文紧紧围绕模分复用系统中的模式解复用问题,在深入分析模式传输特性、不同损伤产生机制以及损伤对信号作用机理的基础上,依据矩阵传输理论推导了模分复用系统的传输模型,并据此引入多模盲均衡、多通道盲解卷积等无需数据辅助的盲信号处理技术来补偿模分复用系统中的各种损伤,消除模式信号的卷积性混合,最终实现解复用。本文的主要工作和创新点具体包括以下三个方面:1.针对常用的盲均衡类解复用算法在高阶调制格式时稳态剩余误差较大等问题,提出基于时域多模盲均衡的模式解复用算法。通过对单输入单输出多模盲均衡算法的扩展,构造出适用于模分复用系统的多输入多输出多模盲均衡算法。并通过搭建的模分复用仿真系统分别验证时域多模盲均衡算法对于低阶调制格式和高阶调制格式的解复用性能。结果表明:多模盲均衡算法在调制格式为4QAM、16QAM和64QAM时均能有效实现解复用,且数字信号处理模块中无需加入载波相位恢复算法。此外,针对时域多模盲均衡算法用于长距离模分复用系统时计算复杂度过高的问题,提出多模盲均衡算法的频域实现算法,即频域多模盲均衡算法。为满足输入信号的二倍过采样和分块自适应构建出包含奇偶子滤波器的蝶形均衡器;并利用快速傅里叶变换将时域多模盲均衡算法扩展到频域;推导出频域多模盲均衡算法具体的梯度更新规则。结果表明:频域多模盲均衡算法在保留时域算法解复用性能的基础上,计算复杂度大幅降低,与常用频域最小均方算法的复杂度位于同一个量级。2.针对模分复用系统特殊的信道传输特性,提出基于多通道盲解卷积的模式解复用算法,直接在时域对卷积性混合的输出信号进行盲分离。首先对卷积性混合的输出信号进行时延向量扩展,将复杂的卷积性混合问题转化为相对简单的瞬时混合问题,然后通过成熟高效的独立成分分析算法实现模式信号的分离。结果表明:在低阶调制格式时,多通道盲解卷积算法的解复用性能与常用均衡算法相近,但在高阶调制格式时,其性能要优于均衡算法。此外,针对时域多通道盲解卷积算法计算复杂度高的问题,提出多通道盲解卷积的频域实现方法—频域独立成分分析算法。算法通过离散短时傅里叶变换将时域内的卷积混合简化为频域内的瞬时混合;时频变换后,利用复值独立成分分析算法对各频点频域分量进行分离,并通过排序算法消除由独立成分分析处理引起的次序不确定性;最终通过逆短时傅里叶变换恢复出时域估计信号。复杂度分析结果表明:频域独立成分分析算法大幅降低了解卷积性混合的复杂度,且表现出对系统模式个数不敏感的优势。但是,当系统差分模式群时延增大到一定程度后,算法复杂度急剧上升。3.针对基于频域独立成分分析的模式解复用算法在大差分模式群时延时会出现复杂度陡升的问题,提出改进的连续收敛的频域独立成分分析算法。理论推导了次序不确定性的成因并通过仿真实验验证次序不确定性对解复用性能的具体影响。为避免排序算法所带来的计算负担,依据相邻频点分离矩阵相近这一特性,为各频点分离矩阵初始化提出一种新的方法。该方法使用较低频点收敛后的分离矩阵作为下一个较高频点分离矩阵的初值,以控制下一个相邻较高频点分离矩阵的收敛趋向,保证相邻频点的瞬时分离系统具有连续的收敛方向和一致的信号输出顺序,从而避免了次序不确定性,省去了额外的排序算法。复杂度分析结果表明:改进的连续收敛的频域独立成分分析算法的复杂度要优于频域独立成分分析以及频域最小均方算法。同时,仿真结果表明:改进的连续收敛的频域独立成分分析算法是一种快速高效的解复用算法,其解复用性能与频域独立成分分析和频域最小均方算法相当,但其收敛速度相对于频域独立成分分析算法来说提升54.5%。本文通过对多模盲均衡算法和多通道盲解卷积理论的深入研究,结合模分复用系统自身特殊的信道传输特性,提出了适用于模分复用系统解复用的新算法。所提出的算法在提升系统输出效率、补偿相位噪声、对高阶调制格式的适用性特别是计算复杂度等方面较现有算法均具有一定程度的改善,为进一步研究适用于模分复用系统的解复用方法提供参考。
[Abstract]:In the world the explosive growth of data traffic in the background, the capacity of single mode optical fiber communication system has been more and more close to the nonlinear Shannon limit. As a new technical breakthrough capacity of single mode fiber, mode diversity multiplexing technology through mining mode of this new degree of freedom can increase exponentially the capacity of the communication system. However, in the actual based on the few mode fiber mode division multiplexing system, mode coupling and mode differential group delay will damage the transmission performance deterioration of the mode division multiplexing system, transmission distance mode division multiplexing system. The interaction pattern of group delay in mode coupling and difference, signal in the transmission process of parting multiplexing system than single mode fiber system is more complex system: each output signals are multiplexed input signal and the channel impulse response of linear superposition convolution. Therefore, how to effectively compensate The mode coupling mode and differential group delay damage, solve the multi channel convolution of mixed problem, realize the output signal demultiplexing is the key problem in mode division multiplexing system. This paper focus on the mode multiplexing mode division multiplexing system, in-depth analysis of mode of transmission characteristics, different damage mechanism and damage mechanism of signal on the basis of the model according to transmission mode division multiplexing system derived transfer matrix theory, and based on the introduction of multi modulus blind equalization, blind signal processing technology to all kinds of damage to non data aided multichannel blind deconvolution to compensate for mode division multiplexing system, eliminate the convolution mixed mode signal, finally realizes the demultiplexing. The main work and innovation points of this paper include the following three aspects: 1. commonly used for blind equalization solutions multiplexing algorithm in high order modulation format steady residual error and so on This time, the mode of multi modulus blind equalization algorithm based on demultiplexing. By extending the single input and single output and multi modulus blind equalization algorithm, constructed for multi input multi output mode division multiplexing system. Then multi modulus blind equalization algorithm to verify the time domain multi modulus blind equalization algorithm for demultiplexing performance of low order and high order modulation modulation format format. By building the mode multiplexing simulation system the results show that the multi modulus blind equalization algorithm in the modulation format for the 4QAM, can effectively achieve the demultiplexing 16QAM and 64QAM, and the digital signal processing without the addition of carrier phase recovery algorithm module. In addition, according to the time domain multi modulus blind equalization algorithm for long distance mode multiplexing when the computational complexity problem, proposed frequency domain multi modulus blind equalization algorithm algorithm, namely frequency domain multi modulus blind equalization algorithm. In order to meet the two times of the input signal over sampling and self block To construct a butterfly parity equalizer consists of sub filters; and by using the fast Fourier transform of the time domain multi modulus blind equalization algorithm is extended to the frequency domain; gradient derived frequency domain multi modulus blind equalization algorithm specific update rules. The results show that the multimode frequency domain blind equalization algorithm based on time domain algorithm retains the demultiplexing performance, computational complexity greatly reduced with the commonly used frequency domain, LMS algorithm complexity in the same order of magnitude for.2. channel transmission characteristics of mode division multiplexing system the proposed multi-channel blind deconvolution algorithm based on mode demultiplexing, directly in the time domain output signal of convolution mixture for blind source separation. The first output signal of the convolution of mixed delay vector expansion, convolution mixed complex problems into instantaneous mixing problem is relatively simple, and then through the independent component analysis algorithm is mature and efficient The separation mode signal. The results showed that: in the low order modulation format, demultiplexing performance with the common equalization algorithm for multichannel blind deconvolution are similar, but in the high order modulation format, its performance is superior to the balancing algorithm. In addition, the time-domain multi-channel blind deconvolution algorithm has high computational complexity problem. Implementation method of frequency domain independent component analysis algorithm, frequency domain multi-channel blind deconvolution algorithm. Through the discrete short-time Fourier transform the time domain convolution is simplified as the instantaneous mixture in frequency domain; time-frequency transform, using complex independent component analysis algorithm is used to separate the frequency domain components, and through the sorting algorithm to eliminate by independent component analysis caused by the order of uncertainty; finally through inverse Fourier transform to recover time-domain estimation signal. The complexity analysis results show that the frequency domain independent component analysis algorithm Understand the convolution mixture significantly reduced complexity and performance to a number of system mode is not sensitive to the advantage. However, when the system differential group delay increases to a certain extent, the complexity of the algorithm is.3. for a sharp rise in frequency domain independent component analysis model based on the demultiplexing algorithm model group in differential time delay there will be a steep rise in the complexity of the problem, the frequency domain independent component analysis is proposed for continuous convergence of improved algorithm. Theoretical derivation of order uncertainty and verified by simulation experiment order uncertainty on the de multiplexing performance of specific effects. In order to avoid the computational burden sorting algorithm caused by adjacent frequency points according to the separation matrix similar to this for the characteristics of the frequency separation matrix initialization of a new method is put forward. This method uses a separation matrix of initial frequency point after the convergence as a relatively high frequency point of the separation matrix, Convergence with higher frequency point adjacent the separation matrix under the control of a trend, ensure the instantaneous separation system of adjacent frequency with the convergence direction of continuous and consistent signal output in order to avoid the order of uncertainty, without additional complexity. Sorting algorithm analysis results show that the frequency domain independent component analysis of the continuous improvement of convergence the complexity of the algorithm is better than the frequency domain independent component analysis and frequency domain LMS algorithm. At the same time, the simulation results show that the frequency domain independent component analysis algorithm for improved convergence is a fast and efficient solution for complex algorithm, the demultiplexing performance and frequency domain independent component analysis and frequency domain LMS algorithm, but its convergence speed relative to the frequency domain independent component analysis algorithm for 54.5%. based on the deep research on the multi modulus blind equalization algorithm and multi-channel blind deconvolution theory, combination mode Multiplexing system special channel transmission characteristics, proposes a new algorithm for mode division multiplexing demultiplexing system. The proposed algorithm in improving the system output efficiency, phase noise compensation, suitable for high order modulation format especially computing complexity compared with the existing algorithms are improved to a certain extent, provide the reference for the further research of application of demultiplexing method in mode division multiplexing system.

【学位授予单位】：吉林大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN929.11

【相似文献】